| Objectives:From the perspective of the sales data of each store,this paper analyzes the overall profit level of the chain brand,comprehensively considers the profit status and stability of the convenience store,establishes a set of comprehensive rationality evaluation rules for the chain convenience store,helps the decision-maker to effectively understand the overall operation status of the enterprise,puts forward corresponding suggestions to improve the operation,and provides a set for the domestic chain convenience store operation research new ideas and frame for reference.Methods:The data of this study is from the daily sales database of Y chain convenience store in X city.The data of the enterprise from June 1,2017 to December 19,2018 are intercepted,with a total of more than 80.2 million detailed single records.The daily revenue,gross profit(hereinafter referred to as profit),cost,sales volume,gross profit rate and passenger flow of each store are extracted,and the stores and sales records that meet the inclusion criteria are retained as the original data set,with a total of 152 stores and 51381 observations.This paper's insight into the overall operation of the enterprise is mainly through the evaluation of the profitability of each store,so the selected evaluation index is profit,and the stability of operation needs to be considered,even if the same brand has a large difference in its daily average profit,so the coefficient of variation is selected to measure.Due to the special location of many stores,there will be a large gap in revenue on weekdays and weekends.Therefore,the profit of each store will be grouped according to whether it is weekend or not,and further difference analysis will be made.The results show that 66.45% of the total daily profit of the store on weekdays and weekends is statistically significant,so the evaluation indexes are the average daily profit and its coefficient of variation on weekdays,the average daily profit and its coefficient of variation on weekends,respectively.The weight of comprehensive evaluation is combined with subjective weight and objective weight.The comprehensive ranking adopts TOPSIS method,rank sum ratio method and functional coefficient method.By using C value,a comprehensive index in TOPSIS evaluation results,the clustering of ordered samples can further achieve the goal of grading.Through the above method,not only the comprehensive operation strength ranking of 152 stores can be realized,but also their grades can be marked.On the one hand,the enterprise managers are concerned about how the value of these evaluation indexes determines the grade,on the other hand,whether they can make an effective prediction of the profit level of new stores through modeling,so we use decision tree and random forest to classify and learn the evaluation indexes and grade marks.Results:1 For the weight of evaluation index,the combination of subjective weight and entropy weight is used to calculate the average daily profit and its coefficient of variation on working days,and the average daily profit and its coefficient of variation on weekends are 0.53,0.18,0.2 and 0.09,respectively,which are consistent with the expected results,indicating that the weight setting is reasonable.2 According to the results of three comprehensive evaluation methods,TOPSIS method,rank sum ratio method and functional coefficient method,33.55% of the store ranking differences are between [0,5],34.87% of the store ranking differences are between(5,10],and the cumulative 87% of the store ranking differences fluctuate between [0,15].The result of the consistency test was chi square value of 5.657,P = 0.059(> 0.05).The difference was not statistically significant.It can be considered that the sorting results of the three methods were basically the same.3 The C value in the TOPSIS ranking result is a comprehensive index,which indicates the closeness of all evaluation objects to the optimal scheme.It is the value calculated by comprehensively considering the average daily profit and variation coefficient of each subsidiary store on weekdays and weekends.The results show that 152 stores can be roughly divided into 5 grades.Among them,the first level is extremely excellent store: 2 stores,accounting for 1.32%;the second level is relatively excellent store: 23 stores,accounting for 15.13%;the third level is good store: 34 stores,accounting for 22.37%;the fourth level is medium store: 50 stores,accounting for 32.89%;the fifth level is poor store: 43 stores,accounting for 28.29%.4 Compare the average profit of multiple samples among different grades,and use ANOVA to compare the average profit of working days.The test result is F=296.54,P<0.0001,which indicates that the total average profit of each grade in the working day group is different.SNK test is used to compare the results between the two groups,and the results show that there are differences between any two groups.In the weekend group,Kruskal Wallis rank sum test was used.Chi square value of test results was 118.49,P<0.0001.It can be explained that the overall distribution position of profit between different grades is different.DSCF method is used to compare the test results.The results show that there are differences between any two groups.5 152 stores were randomly divided into training samples and verification samples according to the proportion of 7:3.The training samples were used for learning,while the verification samples were used for prediction.The prediction accuracy of decision tree model is 76.09%,while that of random forest model is 80.43%.The prediction effect is ideal.Conclusions:1 The evaluation index constructed can reflect the comprehensive profit level of each store of the chain brand,and the ranking results given by the three comprehensive evaluations can be considered basically the same.2 The 152 stores can be divided into five levels by the ordered sample clustering,and the difference of profit level between each level is statistically significant.However,the brand has only 38.82% of its stores are in good condition.The overall business situation is to be improved.3 The prediction accuracy of two kinds of machine learning is considerable,which has a certain promotion significance for the profit level judgment of new stores. |
PDF Full Text Request